Detection and analysis of partial discharges in an electrical signal created by pulse width modulation (PWM)

Abstract

A method for detecting and analyzing a partial discharge (PD) in an electrical signal (1) to be analyzed, generated by PWM and originating from electrical equipment (10) in an aeronautical environment, comprising a detection phase (S1), including a detection step (E10) using a sensor (91) of the electrical signal to be analyzed (1), and a recognition step (E1) of a possible partial discharge by comparing this electrical signal with at least one signature signal (2) representative of a partial discharge (PD), and an analysis phase (S2), including a decomposition step (E4) of the electrical signal (1) into a sequence of patterns, and an assignment step (E5) of the detected partial discharges (PD) to these patterns. Figure for the abstract: Fig. 1

Description

Title of the invention: Detection and analysis of partial discharges in an electrical signal created by pulse width modulation (PWM) FIELD OF INVENTION

[0001] The present invention relates to the identification of partial discharges under a voltage synthesized by pulse-width modulation, PWM. It is particularly applicable to assisting in the identification of a type of fault in an electrical chain, and in an aeronautical environment.

[0002] A partial discharge, referred to as DP in the following, is a localized electrical discharge that only partially short-circuits the insulating gap separating conductors or electrodes.

[0003] The presence of these discharges leads to accelerated degradation of the insulation, whether liquid, through oxidation, or solid, through erosion, and can lead to significant reliability problems. Beyond premature wear of the onboard equipment, the recurrence of partial discharges can cause failures that can be critical in the case of an aircraft in flight.

[0004] Until recently, the voltage levels used in aeronautics were not high enough to cause significant concern regarding the existence of these partial discharges, which were unlikely to occur and therefore caused little damage to the equipment. Commonly used voltage levels were, for example, in the range of 230 to 400 volts maximum.

[0005] Furthermore, various restrictions on carbon emissions have been, are being, or will be adopted by various States. In particular, an ambitious standard applies both to new types of aircraft and to those currently in operation, requiring the implementation of technological solutions to bring them into compliance with current regulations. Civil aviation has been mobilizing for several years now to contribute to the fight against climate change.

[0006] Technological research efforts have already led to very significant improvements in the environmental performance of aircraft. The Applicant takes into account the factors impacting all phases of design and development in order to obtain aeronautical components and products that are less energy-intensive, more environmentally friendly, and whose integration and use in civil aviation have moderate environmental impacts, with the aim of improving the energy efficiency of aircraft.

[0007] Consequently, the Applicant is constantly working to reduce its climate impact by using methods and operating virtuous development and manufacturing processes that minimize greenhouse gas emissions to the minimum possible in order to reduce the environmental footprint of its activity.

[0008] This sustained research and development work focuses in particular on the use of electrical technologies to provide propulsion.

[0009] This trend towards hybridization and / or electrification of propulsion systems leads to an increasing demand for electrical energy and consequently to higher voltage levels. Thus, the voltage levels used on aircraft electrical systems can now exceed 400 volts and even reach kilovolts. This voltage increase, combined with severe pressure and temperature conditions, increases the risk of these partial discharges occurring.

[0010] In addition, in certain applications, voltages are generated by pulse width modulation (PWM); these voltages are characterized by steep rising and falling edges (high dV / dt derivative), further increasing the probability of partial discharges occurring.

[0011] Wherever possible, the various components are designed to prevent the occurrence of such partial discharges. However, it is impossible to completely prevent this phenomenon in the long term, due to the natural wear of the components which, in the aeronautical field, are subjected to severe stresses, particularly in terms of temperature and pressure.

[0012] Due to the increased risk of partial discharges occurring in harsh environments, and their significant impact on the reliability of onboard equipment, it is necessary to be able to reliably detect and recognize these partial discharges as soon as they occur, without requiring excessive computing power. This last point is all the more important in an embedded system and in an aeronautical context where computing resources may be limited.

[0013] Moreover, while under sinusoidal voltage partial discharge detection is relatively simple, under impulse stress or generated by Pulse Width Modulation (PWM), the noise generated tends to be superimposed on the partial discharge signals, thus complicating the detection of partial discharges.

[0014] Since the discharges have very low charge values, complex and robust measuring devices must be implemented.

[0015] Detection methods exist but are not very simple to implement because they require suitable electronics and high requirements in terms of acquisition frequency, for example. Furthermore, verification by an expert is often necessary to confirm or not the possible presence of partial discharge.

[0016] An object of the invention is therefore to improve current proposals of the state of the art by enabling, in particular, the detection and then the identification (i.e. the characterization) of a partial discharge in an electrical voltage signal, in particular when it is generated by pulse width modulation, PWM.

[0017] This characterization may include in particular a phase-amplitude-frequency analysis, of the PRPD type (“Phase Resolved Partial Discharge” in English, for “phase-resolved partial discharge”). Summary of the invention

[0018] In order to overcome the shortcomings of the prior art concerning the detection of partial discharges, particularly with regard to voltage signals generated by pulse-width modulation (PWM), according to a first aspect, the present invention can be implemented by a method for detecting and analyzing a partial discharge, DP, in an electrical signal to be analyzed originating from electrical equipment in an aeronautical environment, comprising

[0019] - a detection phase, comprising a detection step using a sensor of the electrical signal to be analyzed generated by pulse-width modulation (PWM), and a step of recognizing a possible partial discharge by comparing said electrical signal to be analyzed with at least one signature signal representative of a partial discharge, and,

[0020] - an analysis phase comprising a step of decomposing said electrical signal to analyze into a sequence of patterns, and a step of assigning the detected partial discharges to said patterns.

[0021] According to preferred embodiments, the invention comprises one or more of the following features which can be used separately or in partial combination with each other or in total combination with each other: - the process further includes a step of associating each of said reasons with information relating to the partial discharges previously assigned to said reason; - said association step includes the superposition of a plurality of partial discharges detected for a plurality of periods of said electrical signal to be analyzed for the same pattern; - said decomposition step includes determining a number of patterns in said electrical signal to be analyzed from parameters of a control signal used to generate said electrical signal by pulse width modulation; - said number of patterns is determined as a function of the frequency of a modulating signal and the frequency of a carrier. - said decomposition step includes an identification of patterns by comparison with a signature pattern; - an amplitude-phase-frequency pattern, of the PRPD type, is associated with the patterns; - said recognition step includes a step of calculating at least one value of a distance parameter between said electrical signal to be analyzed and said at least one signature signal, said distance parameter being a function of a difference between said electrical signal to be analyzed and said at least one signature signal, and a step of comparing the value of said distance parameter with a detection threshold, and of detecting said possible partial discharge as a function of a result of said comparison.

[0022] Another object of the invention relates to a computer program comprising instructions to implement a process as previously described when said instructions are executed by a processor of a detection and analysis device.

[0023] Another object of the invention relates to a detection and analysis device, adapted to implement the steps of the process as previously described.

[0024] Another object of the invention relates to an aircraft comprising at least one such detection and analysis device.

[0025] Other features and advantages of the invention will become apparent from the following description of a preferred embodiment of the invention, given by way of example and with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE FIGURES

[0026] The attached drawings illustrate the invention:

[0027] Figure 1 shows a flowchart of the steps of a method for detecting and recognizing partial discharges according to one embodiment of the invention.

[0028] Figure 2 illustrates the technical principle behind the recognition of partial discharges according to embodiments of the invention.

[0029] Figure 3 illustrates a partial discharge detection system for obtaining signatures according to an embodiment of the present invention.

[0030] Figure 4 illustrates a possible construction of a PWM electrical signal.

[0031] Figure 5 illustrates an example of the decomposition of an electrical signal into two patterns,

[0032] Figures 6A-6D illustrate simplified examples of PRPD schemes in the context of sinusoidal signals,

[0033] Fig. 7 illustrates a simplified example of PRPD type schemes for a PWM type electrical signal.

[0034] DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

[0035] An electrical insulator, by preventing the passage of electric current, allows the proper functioning of electrical systems. In power electronics, these materials are found either in passive components, where the focus is on exploiting the dielectric properties of the material (and its energy storage capabilities), or as insulation. In electrical engineering, the primary function is to insulate parts held at different potentials.

[0036] Insulating materials, whether solid, liquid or gaseous, are often the weak link in electrical systems; in particular, beyond a certain voltage, the phenomenon of Partial Discharges (PD) and its consequences (such as the appearance of short circuits or fault arcs) are likely to occur.

[0037] Until now, the existence of these partial discharges had been largely overlooked in the design of aircraft equipment, given the low voltage levels used. However, particularly in the aeronautical field, the hybridization and electrification of high-power propulsion systems leads to an increase in operating voltages; consequently, no current system can claim to be either free from or resistant to partial discharges.

[0038] Hereafter, partial discharge (PD) will be understood to mean a localized electrical discharge generated, under the effect of a high voltage or voltage variation, in an insulating gap separating conductors. Furthermore, electrical network will be understood to mean any type of network that can be found in an aeronautical environment, for example in any aircraft: airplane or helicopter, etc.

[0039] The [Fig. 1] is the flowchart of the steps of a process for detecting and recognizing partial discharges DP, the main steps of which are illustrated by the [Fig. 2].

[0040] The proposed method comprises a detection phase SI, and an analysis phase S2. This method is implemented here by computer.

[0041] In the detection phase SI, the method includes a detection step E10 of an electrical signal to be analyzed 1.

[0042] Pulse-width modulation (PWM) is a commonly used technique for synthesizing pseudo-analog signals using digital (on / off, 1 or 0) circuits, or more generally, discrete-state circuits. The general principle is that by applying a rapid succession of discrete states with carefully chosen duration ratios, any intermediate value can be obtained by only considering the average value of the signal.

[0043] Various techniques exist for generating an electrical signal by PWM.

[0044] A classic technique is the so-called intersective method.

[0045] With reference to [Fig.4], this process consists of making a comparison, at each instant, between the values ​​of a modulating signal 42 and a carrier 41, and assigning a discrete value to the modulated electrical signal 1 according to the result of this comparison.

[0046] The carrier can be a triangular signal, as in [Fig.4], but other waveforms (sinusoids...) are also possible.

[0047] Fig. 4 illustrates a possible construction of an electrical PWM signal from the triangular carrier 41 of the modulating signal 42, here sinusoidal.

[0048] When the modulating signal 42 is greater than the carrier 41, as in zone 43, the output electrical signal 1 takes a first discrete value, corresponding, for example, to a logic "1". Otherwise, as in zone 44, for example, the electrical signal 1 takes a second discrete value, corresponding, for example, to a logic "0". In this way, an output signal is obtained that forms an alternation of discrete values ​​(typically two) depending on the comparison of the modulating signal 42 and the carrier 41.

[0049] The carrier 41 can also be sinusoidal, or sawtooth, etc.

[0050] Another method of generating a PWM signal is the 3rd order harmonic injection method. This method was first described in Buja's article, "Improvement of puise width modulation techniques", Archiv für Elektrotechnik, vol. 57, no. 5, 1975, pp. 281-289.

[0051] This method can be seen as a variant of the intersective method, but includes a prior step of modifying the modulating signal by adding a factor in y • sin(3mi), i.e. by inserting a 3rd order harmonic into the modulating signal 42. The value m is a parameter of the process and t represents time.

[0052] Another method for generating a PWM signal is the space vector modulation (SVM) method. It was first described in G. Pfaff, et al., “Design and experimental results of a brushless AC servo drive”, IEEE Transactions on Industry Applications, vol. IA-20, no. 4, pp. 814-21, July / August 1984, and is also widely described in the scientific literature on the subject, for example in reference works such as MP Kazmierkowski; R. Krishnan & F. Blaabjerg (2002). “Control in Power Electronics: Selected Problems”, San Diego: Academie Press. ISBN 978-0-12-402772-5.

[0053] Another technique is the pre-calculated method, also called offline PWM or "Optimal Pulse Pattern" (OPP). The pattern of the output electrical signal, 1, is predetermined (offline) and stored in tables which are then read in real time. The method has, for example, been described in AD Birda, J. Reuss, and C. Hackl, “Synchronous optimalpulse-width modulation with differently modulated waveform symmetry properties for feeding synchronous motor with high magnetic anisotropy”, in 2017 19th European Conference on Power Electronics and Applications (EPE' 17 ECCE Europe), pages 1-10, Sept 2017, doi:10.23919 / EPE17ECCEEurope.2017.8098963.

[0054] Another technique is full-wave control. In this type of operation, the switches used to generate the PWM signals operate at the frequency of the output electrical quantities. The conduction time of a switch is T / 2, where T is the electrical period. The generated signal is then a periodic square wave with period T.

[0055] Other methods have been proposed for generating electrical signals by pulse-width modulation (PWM) and are described in the technical literature. As will be seen later, the described method can be adapted to PWM signals generated by different methods.

[0056] Once detected, the electrical signal to be analyzed 1 is advantageously preprocessed. The detection step is accompanied by filtering Eli of the electrical signal to be analyzed 1 using a filter. The filter is preferably a high-pass filter, in order to minimize the noise level present in the electrical signal to be analyzed 1. The cutoff frequency of the high-pass filter is on the order of hundreds of MHz and above, for example around approximately 300 MHz.

[0057] The process then includes an acquisition step E12 of an electrical signal to be analyzed 1. The electrical signal to be analyzed 1 is preferably acquired by means of the acquisition device (described below).

[0058] Once the electrical signal to be analyzed 1 is processed, a recognition step, El, of a possible partial discharge is carried out. This step can be based on a comparison between this electrical signal to be analyzed and one or more signature signals, 2, representative of a partial discharge, DP.

[0059] Different embodiments of this El recognition step can be implemented within the framework of the proposed process.

[0060] According to one embodiment, this recognition step El includes a calculation step E2 of a value of a distance parameter 3. The distance parameter 3 is a function of a gap between two signals or portions of signals.

[0061] The calculation E2 of a value of the distance parameter 3 is performed between the electrical signal to be analyzed 1 and a signature signal 2. The signature signal 2 is a signal representative of the signal of a partial discharge DP. The value of the distance parameter 3 is therefore calculated between the electrical signal to be analyzed 1 and a signal representative of a partial discharge DP.

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072] According to one implementation method, the value of the distance parameter 3 is calculated as a normalized Euclidean distance. That is, for each point of the electrical signal to be analyzed 1, whose magnitude is centered and reduced, its distance from the corresponding point of the signature signal 2, whose magnitude is also centered and reduced, is determined. In other words, the value of the distance parameter 3 is the result of the equation: With N being the number of points in the signature signal 2, x a value of a point in the signature signal 2, and y a value of a point in the electrical signal to be analyzed 1. The values ​​of points x and y are advantageously voltages. Preferably, the values ​​are centered and scaled to obtain values ​​independent of the unit or scale chosen and having the same mean and dispersion. In another implementation, the value of the distance parameter 3 is calculated using a dynamic time warping function. Such a function allows the value of the distance parameter 3 between the electrical signal to be analyzed 1 and the signature signal 2 to be calculated more robustly with respect to temporal and amplitude dilation / contraction. According to another implementation method, it is possible to calculate the value of the distance parameter 3 using any function suitable for evaluating a gap between two signals. Furthermore, calculation step E2 advantageously includes calculating a series, 8, of values ​​for the distance parameter 3. The series 8 of values ​​is obtained by implementing the following steps: - a selection E21 of a portion 7 of the electrical signal to be analyzed 1 in an analysis window 6, the analysis window 6 having a predefined width 61; - an E22 calculation of a value of the distance parameter 3 between the portion 7 selected by the analysis window 6 of the electrical signal to be analyzed 1 and the signature signal 2; - the E23 offset of the analysis window 6 by a predefined step 62; and - the E24 renewal of these selection steps E21, calculation E22 and shift E23 in order to obtain the series 8 of values ​​of the distance parameter 3. In other words, the analysis window 6, which selects a portion 7 of the electrical signal to be analyzed 1, is moved temporally along the electrical signal to be analyzed 1, and for each of its positions, the calculation E22 of the distance parameter 3 is performed. In this way, the series 8 of values ​​of the distance parameter 3 represents the gap between the electrical signal to be analyzed and the signature signal for each of the portions 7 of this electrical signal to be analyzed 1.

[0073] Advantageously, the width 61 of the analysis window 6 represents a time interval corresponding to the duration of the partial discharge DP represented on the signature signal 2.

[0074] Advantageously, the step 62 of the shift of the analysis window 6 along the time scale of the electrical signal to be analyzed 1 is equal to an acquisition period of the electrical signal to be analyzed 1. It is however possible to choose a step 62 different from this acquisition period such as for example a multiple of the acquisition period or any other time period.

[0075] Moreover, it is advantageously possible to calculate E22 a value of a distance parameter 3 between a portion 7 of the electrical signal to be analyzed 1 and a plurality of signature signals 2 representative of different types of partial discharges DP.

[0076] According to one embodiment, the plurality of signature signals 2 is derived from a database. A detection device C, described below, allows for the generation, in an optional step E0, of a database comprising a plurality of signature signals 2 representative of partial discharges DP. It is indeed possible to identify the signature signals and record them in the database prior to the implementation of the other steps of the detection and recognition process. The signatures of signals 2 can evolve according to the conditions in which the electrical network 10 finds itself during its use. It is therefore possible to generate, in the laboratory, signatures of signals 2 representative of partial discharges DP as they would be under these conditions.In order to be able to monitor the state of the insulation of the electrical network 10, it is advantageous to implement the detection and recognition process including the use of the database 10 comprising a plurality of signature signals 2 representative of partial discharges DP. .

[0077] Furthermore, different signature signals 2 can be collected for the same type of partial discharge, depending in particular on the location where the measurement is carried out. Thus, the same partial discharge phenomenon will generate a different signal depending on whether it is measured by probe 91 or probe 92, both located on the same line 10 but at a certain distance from each other (for example 1 or 2 meters), on [Fig. 3], described later.

[0078] After the calculation step E2, the detection and recognition process includes a comparison step E3. During this step, the value of the distance parameter 3 from the calculation step E2 is compared with a detection threshold 4. The comparison E3 is designed to detect a possible partial discharge DP based on the results obtained, and therefore on the distance between the electrical signal to be analyzed 1 and the signature signal 2. Indeed, the lower the value of the distance parameter 3, the smaller the difference between the electrical signal to be analyzed 1 and the signature signal 2. A small difference between the electrical signal to be analyzed 1 and the signature signal 2 implies a high similarity, and the detection of a partial discharge DP is then likely.

[0079] In the embodiment in which a series 8 of values ​​of the distance parameter 3 is calculated, the comparison step E3 compares the series 8 to the detection threshold 4 in order to detect one or more possible partial discharges DP and their position in the electrical signal to be analyzed 1.

[0080] The detection threshold 4 is advantageously determined as a function of the mean of the series 8 of values ​​of the distance parameter 3 from which n times the value of the standard deviation of the series 8 of values ​​of the distance parameter 3 is subtracted. The number n is a real number, advantageously an integer greater than or equal to 1.

[0081] According to one embodiment, the detection and recognition method described above can be coupled with other methods to ensure even more precise recognition. For example, the method can be coupled with a partial discharge DP recognition method implementing a wavelet transform of the electrical signal to be analyzed 1.

[0082] Fig. 3 schematically illustrates a device C allowing the detection, in the laboratory, of signature signals 2, enabling the generation of the database of signature signals 2, E0, and / or obtaining the electrical signal to be analyzed 1 in order to implement the detection and recognition process as described above.

[0083] Furthermore, a source S of an electrical signal to be analyzed and a motor M are represented, the motor M being able to be replaced by any device consuming electrical energy.

[0084] The source is adapted to generate an electrical signal by pulse-width modulation. To do this, it can receive a command from a control device, not shown, which forms a modulating signal. It can then modulate a carrier signal (typically a pure sine wave) to form a PWM-type electrical signal.

[0085] The device C is, for example, a test bench comprising a data acquisition device 9 and a sensor 91. The sensor 9 is, for example, a capacitive coupling type sensor 91. The sensor 91 advantageously comprises a metallic tip that touches a portion 11 of an electrical network 10. The portion 11 of the electrical network 10 is advantageously coated with a layer 12 of copper to amplify the detected electrical signal.

[0086] However, other sensors besides capacitive sensors can also be used to implement step E10 of detecting an electrical signal.

[0087] The device C allows both to generate in advance E0 the database, but also to obtain the electrical signal to be analyzed 1. To this end, the sensor 9 acquires the electrical signal, for example in millivolts, image of an electrical quantity varying in the electrical network 10, for example the electric current.

[0088] An Eli filtering process, for example filtering using a high-pass filter, is applied to the electrical signal in order to reduce the noise level as much as possible. The cutoff frequency of the filter used is advantageously around one hundred megahertz or higher, for example around approximately 300 MHz. Finally, the acquisition step E12 is necessary to obtain the electrical signal to be analyzed 1 or the signature signal 2.

[0089] Once detected, filtered and acquired, the signal is either recorded in the database if it is a measurement intended to enrich E0 the database in the laboratory, or used as an electrical signal to be analyzed 1 to implement the detection and recognition process as described above.

[0090] An analysis phase S2 of the proposed process includes a decomposition step E4 of the electrical signal to be analyzed 1 into a sequence of patterns.

[0091] Indeed, the generation of the electrical signal by pulse-width modulation (PWM) is carried out from two periodic signals, a modulating signal and a carrier. Necessarily, the resulting signal 1 is also periodic.

[0092] However, its period may be longer than that of the input signals and may not correspond to the period of the sinusoid corresponding to the PWM signal. Therefore, at each period of the sinusoid, a different PWM pattern may be observed.

[0093] Figure 5 illustrates a decomposition of an electrical signal 1 into two motifs M1, M2. These two motifs follow one another and thus compose the signal 1.

[0094] Several methods are possible for determining the patterns into which the electrical signal 1 can be decomposed. These methods may depend on the method that was used to generate the PWM signal.

[0095] According to one embodiment, a frequency-domain method is used. This method is based on prior knowledge of the carrier frequency fp and the modulating signal frequency fm. The number of patterns that can be generated can then be deduced. This information is sufficient to fully characterize the output signal since it is necessarily a succession of these different patterns. Knowing the number N of patterns, it can therefore be deduced that the output signal is a modulo N succession of patterns mb m2...mN.

[0096] In the case of an electrical signal 1 generated by the intersection method, this number of patterns can be determined by the expression:

[0097] LCM(fP-fm) f

[0098] The LCM() notation indicating the least common multiple.

[0099] In the case of an electrical signal generated by other methods, this number of patterns can be determined in different ways.

[0100] In the case of an electrical signal generated by harmonic injection or by SVM, the frequency-domain method can also be used. Indeed, these methods are derivatives of the intersection method but with a modulating waveform of a different shape. From prior knowledge of the frequencies of the carrier wave fp and the modulating signal fm, one can then deduce the number of patterns that can be generated.

[0101] Regarding the pre-calculated method, the waveforms are preferably determined a priori by means of a calculation of switching angles allowing optimization of the frequency spectrum of the generated signal. Knowing the value of these angles makes it possible to determine the number of patterns of the input electrical signal.

[0102] Finally, in the case of full-wave control, the switches operate at the frequency of the output electrical quantities. The pattern of the generated PWM signal is therefore unique and is a square wave with period T, where T is the electrical period of the signal.

[0103] According to another embodiment, the patterns composing the electrical signal 1 can be determined by identifying these patterns within the signal.

[0104] In particular, an approach similar to that previously described for the detection and recognition of a partial discharge in the signal can be used. Indeed, in the same way as before, a dictionary of signature patterns can be constructed, and a value of a distance parameter between the electrical signal 1 and the signature patterns can then be calculated, this value compared with a threshold, and then a pattern identified based on this comparison.

[0105] The analysis phase S2 of the proposed process then includes an assignment step E5 of the detected partial discharges DP to the previously identified patterns.

[0106] Indeed, since the electrical signal 1 is decomposed into a succession of patterns, each detected partial discharge can be positioned temporally with respect to a pattern resulting from the decomposition of the signal. This DP / pattern assignment can therefore be easily performed.

[0107] Subsequently or in parallel, information relating to these partial discharges can be associated with each pattern: as the electrical signal 1 is analyzed and partial discharges are detected, they can be assigned to the corresponding pattern. The information associated with this pattern is thus progressively enriched with data relating to the detected partial discharges.

[0108] The process therefore includes an association step E6 for each motif, of information relating to the partial discharges (PD) assigned to that motif.

[0109] This information can be structured in different ways. According to a preferred embodiment, this information can include an amplitude-phase-frequency scheme, of the PRPD type (“Phase Resolved Partial Discharge” in English, for “phase-resolved partial discharge”).

[0110] On this type of PRPD scheme, the DPs are represented by a cloud of points characterized by a phase (allowing it to be positioned in relation to the analyzed signal), an amplitude (allowing it to be positioned on the ordinate axis), and a frequency (corresponding to the number of DPs for this phase and this amplitude).

[0111] More generally, the information generated and associated with the patterns can be used to produce PRPD visual diagrams on a human-machine interface, but the information can also be used in other ways, in particular for digital processing without going through a display phase.

[0112] In particular, digital processing can make it possible to automate a diagnostic step allowing the information to be attributed to one or more probable faults causing the partial discharges.

[0113] The recognition of patterns associated with each type of fault can be carried out as in the case of PRPD in the form of a sinusoidal wave (figures 6A-6D), with the objective of constructing a dictionary of patterns / faults in the case of voltage signals generated by PWM.

[0114] With the use of supervised machine learning tools, in particular, it is possible to classify these different patterns and associate them with each type of defect that causes partial discharges.

[0115] Figures 6A-6D illustrate simplified examples of PRPD schemes in the context of sinusoidal signals.

[0116] Each figure represents one period (360°) of the sinusoidal signal, on which is shown the (majority) distribution of the detected partial discharges.

[0117] Fig. 6A illustrates the presence of partial discharges in a cavity created due to an abrasion phenomenon of a semiconducting paint.

[0118] Fig. 6B illustrates the presence of partial discharges in cavities resulting from a delamination phenomenon of insulating tape layers.

[0119] Fig. 6C illustrates the presence of surface discharges at the coil heads of an electrical machine, due to a surface contamination phenomenon.

[0120] Finally, [Fig.6D] illustrates the presence of internal partial discharges in microcavities present in an insulator.

[0121] As can be seen from the various examples in Figures 6A to 6D, the PRPD pattern forms a characteristic signature of a type of defect.

[0122] The proposed method makes it possible to obtain information enabling the generation of such a PRPD scheme within the framework of an electrical signal generated by PWM, that is to say by associating the DPs not with the global signal but with the constituent patterns of this signal.

[0123] This contribution is important because it is the waveforms constituting the patterns themselves that can allow the identification of defects. It is therefore useful to be able to relate the DPs to the different patterns constituting an electrical signal.

[0124] Fig. 7 illustrates a simplified example of PRPD type schemes for a PWM type electrical signal, which can be obtained from the implementation of a detection and analysis process as described above.

[0125] This figure shows two motifs M1, M2 constituting an electrical signal. The information generated in step E6 and associated with each motif therefore makes it possible to generate a PRPD diagram for each of the two motifs.

[0126] This scheme can be obtained by superimposing the detected partial discharges DPs according to their temporal phase and amplitude in the same reference frame as the patterns, and opposite the pattern corresponding to their detection. Preferably, said superposition is carried out in the form of point clouds.

[0127] The distribution of points corresponding to the DPs within the patterns M1, M2 can form a signature of a defect at the origin of the DPs.

[0128] Thus, to summarize, the proposed method makes it possible to perform a PWM-enabled PRPD by indicating the presence of Partial Discharges in the different phases of the analyzed electrical signal. It therefore offers a solution to the problem of analyzing PDPs under PWM enabled voltage.

[0129] The method therefore allows the current or future use of PRPD type schemes for signals generated by pulse width modulation.

[0130] In general, the described process allows: - to reliably and easily trace the PRPD under PWM voltage. - to identify the occurrence of partial discharges (if any). - to enable identification of defects as well as analysis of critical flight phases, for which the frequency of DPs would potentially be higher. - to avoid more complex treatments. - to distinguish all DPs from other high-frequency noise such as that of PWM converters.

[0131] This described method is compatible with all types of electrical systems operating under PWM voltage, regardless of their power and use, provided that one has - sensors capable of providing the signal that contains the signature of the phenomenon and - from the dictionary of signatures to search for.

[0132] The proposed method is also compatible with all the different techniques for generating a PWM signal (intersection PWM, harmonic injection PWM, etc.).

[0133] Of course, the present invention is not limited to the examples and embodiment described and illustrated. In particular, it is susceptible to numerous variations accessible to those skilled in the art.

Claims

Demands

1. Method for detecting and analyzing a partial discharge (PD) in an electrical signal to be analyzed (1) generated by pulse width modulation (PWM), from electrical equipment (10) in an aeronautical environment, comprising - a detection phase (SI), including a detection step (E10) by means of a sensor (91) of the electrical signal to be analyzed (1), and a recognition step (El) of a possible partial discharge by comparison of said electrical signal to be analyzed (1) with at least one signature signal (2) representative of a partial discharge (PD), and, - an analysis phase (S2), including a decomposition step (E4) of said electrical signal to be analyzed (1) into a sequence of patterns, and an assignment step (E5) of the detected partial discharges (PD) to said patterns.

2. A method according to claim 1, further comprising a step of associating each of said motifs with information relating to partial discharges (PD) previously assigned to said motif.

3. Method according to the preceding claim, wherein said information comprises an amplitude-phase-frequency scheme, of the PRPD type.

4. Method according to claim 3, wherein said amplitude-phase-frequency scheme is obtained by superimposing a plurality of partial discharges detected for a plurality of periods of said electrical signal to be analyzed (1) for the same pattern.

5. A method according to any one of the preceding claims, wherein said decomposition step (E4) comprises determining a number of patterns in said electrical signal to be analyzed (1) from parameters of a control signal used to generate said electrical signal by pulse width modulation, said number of patterns being preferably determined as a function of the frequency of a modulating signal and the frequency of a carrier.

6. A method according to any one of claims 1 to 4, wherein said decomposition step (E4) comprises pattern identification by comparison with a signature pattern.

7. A method according to any one of the preceding claims, wherein said recognition step (E1) comprises a calculation step (E2) of at least one value of a distance parameter (3) between said electrical signal to be analyzed (1) and said at least one signature signal (2), said distance parameter being a function of a difference between said electrical signal to be analyzed (1) and said at least one signature signal (2), and a comparison step (E3) of the value of said distance parameter (3) with a detection threshold (4), and of detection of said possible partial discharge (PD) as a function of a result of said comparison.

8. Computer program comprising instructions to implement a method according to any one of the preceding claims when said instructions are executed by a processor of a detection and analysis device (C).

9. Detection and analysis device (C), comprising at least one sensor (91) and an acquisition device (9) adapted to, in collaboration with a database, implement the steps of the process according to any one of claims 1 to 7.

10. Aircraft comprising at least one device according to the preceding claim.

Images